baodq06/memocall
š š š šŖ š§ - Recall past Claude Code conversations from any project. Lists and searches your sessions, then loads one on demand as clean, compact Markdown ā with outline, turn-range, and in-session search for large sessions. Read-only over local /.claude transcripts; collapses tool calls to stay under the context cap. npx -y memocall
Quick Install
{
"mcpServers": {
"baodq06-memocall": {
"command": "npx",
"args": [
"-y",
"baodq06-memocall"
]
}
}
}Using an AI coding agent (Claude Code, Cursor, etc.)? Copy a ready-made prompt that tells it to fetch the setup instructions and install this server for you.
Documentation Overview
MemoCall
An MCP server that lets Claude Code recall the context of your past conversations ā from any project ā on demand.
Claude Code can resume a session, but getting another conversation's context into the one you're in is a chore ā copy-pasting exports or digging through raw transcript files by hand ā and resume only sees the current directory anyway. MemoCall turns it into a single ask: "what was I working on?" or "load the session where we set up X", and it pulls that transcript in as clean, compact Markdown ā even from a totally different project.
It's read-only. Claude Code already records every session to ~/.claude/projects/<project>/<id>.jsonl; this server just reads those files, strips the noise, collapses tool calls, and hands back a readable transcript. It never writes, hooks, or touches a live session.
What you get
Five tools, available in every session once installed. Claude picks the right one from the request:
| Tool | What it does |
|---|---|
list_sessions | Your recent conversations across all projects, grouped by directory. "What was I working on?" |
search_sessions | Find a past conversation by keyword (matches titles + first messages + paths). |
load_session | Load one conversation as compact Markdown ā by id or query. Optionally a turn window via turns. |
session_outline | A cheap map of one conversation: the numbered list of your prompts. Great for huge sessions. |
search_in_session | Return only the turns within one conversation that match a keyword. |
You don't call these directly ā you talk normally and Claude reaches for them:
You: what sessions have i worked on recently?
You: load the one where we set up the license invitation system
You: which session did we debug the keychain SIGKILL thing in?
How the transcript is rendered
Raw transcripts are a verbose event log (one big file hit 11 MB). memocall reduces each turn to the essentials, Conductor-style:
**You:** right now i need to set up a system to invite prospects...
ā³ 12 tool calls (Bash Ć5, Read Ć4, Edit Ć3)
**Claude:** I've set up the invitation flow. Key decisions: ...
Tool-call outputs are elided (the big token win, and a privacy win ā see below). Abandoned/edited message branches are dropped so you get the conversation as it actually played out. A token budget keeps even an 11 MB session well under Claude Code's MCP output cap via middle-out truncation.
Navigating large sessions
A single response can't hold a 1,000-turn session, so for big ones you don't dump ā you navigate:
session_outlineā a numbered map of every prompt (tiny, fits any session).search_in_sessionā jump straight to the turns about a topic, orload_sessionwithturns: "300-340"to pull an exact window (turn numbers come from the outline).
So you never lose access to the middle of a huge conversation ā load_session alone would middle-out-truncate it, but the outline + range/search tools let Claude reach any part on demand.
Install
Requires the Claude Code CLI (and Node 18+). One line ā npx fetches and runs the published package:
claude mcp add --scope user memocall -- npx -y memocall
--scope user makes it available in every session, in every directory. Restart Claude Code (or open a new session) and ask it to list your sessions. Verify with claude mcp list ā you should see memocall: ā¦ ā Connected.
Or run from source (for development)
git clone https://github.com/baodq06/memocall.git
cd memocall && npm install && npm run build
claude mcp add --scope user memocall -- node "$(pwd)/dist/index.js"
Privacy & security
- Local only. No network, no auth, no telemetry. It reads files under
~/.claude/projects/and nothing else. - Compact mode elides tool outputs, which is where secrets (tokens, keys, env) usually live ā so the default output is much safer than the raw transcript.
format: "full"includes brief tool inputs and may surface sensitive strings. Use it deliberately.- Transcripts can contain secrets regardless; treat loaded context as you would the original conversation.
Limitations
- Recall is best-effort: Claude Code deletes transcripts after
cleanupPeriodDays(default 30). - Forked sessions may only contain post-fork turns.
- The transcript format is undocumented and can change between Claude Code versions; all format knowledge is isolated in
src/jsonl.tsso it's a one-file patch if it does.
Development
npm run build # compile TypeScript -> dist/
npm test # unit suite (node:test) on synthetic fixtures
npm run smoke # optional: checks all YOUR real sessions stay under the output cap
node dist/parser.js <file.jsonl> [--full|--outline] [--turns 10-20] [--search "kw"] [--max N] [--think] # test the parser
node test-client.mjs # drive the server over stdio like Claude Code does
npm run inspect # open the MCP Inspector UI
Layout:
src/jsonl.tsā all knowledge of the transcript format (helpers, ordering, classification).src/parser.tsā JSONL ā compact Markdown (the core transform).src/locator.tsā session enumeration + cached metadata index.src/index.tsā the MCP server wiring the three tools.
License
MIT